The invention concerns genetic engineering and screening methods useful for the identification of gene targets for anti-cancer agents.
Cancer is a complex and devastating group of diseases that kills one in five adults in developing countries. Although cancers arise from a wide variety of cells and tissues in the body, there are unifying features of this group of diseases. Cancer is predominantly a genetic disease, resulting from the accumulation of mutations that promote clonal selection of cells that exhibit uncontrolled growth and division. For example, by the time a tumor reaches a palpable size of about one centimeter, it has already undergone about thirty cell doublings, has a mass of approximately one gram, and contains about one billion malignant cells. The result of such uncontrolled growth of tumor cells is the formation of disorganized tissue that compromises the function of normal organs, ultimately threatening the life of the patient. Obviously, methods for prevention, early detection and effective treatment of cancer are of paramount importance.
The past twenty years of research on the mechanistic basis of carcinogenesis have resulted in a revolution in our understanding of the molecular nature of genetic changes that initiate tumor formation. Specific genes have been identified that are frequently mutated in tumor cells, many of which have been grouped into two main classes termed oncogenes and tumor suppressor genes. A few key genes have been identified that are very commonly mutated in a large number of different tumors, such as the oncogene ras and the tumor suppressor genes p53 and Rb. Furthermore, genes that are mutated in tumor cells tend to have functions that cluster in one of the following categories: DNA repair, chromosomal integrity, cell cycle control, growth factor signaling, apoptosis, differentiation, angiogenesis, immune response, and cell migration. Thus, it is clear that there are specific mutations in certain genes that distinguish cancer cells from normal cells.
Despite the fundamental significance of these discoveries, they have not been paralleled by the development of highly selective drugs to treat cancer. This lag in the development of practical therapeutic applications from these discoveries is due to several factors. An ideal chemotherapeutic must selectively kill or block the proliferation of tumor cells without having a deleterious effect on normal growing cells in the body. Most of the genetic alterations found in tumors cells that distinguish them from normal cells are either gain-of-function mutations in oncogenes, which result in increased expression or activity of the gene product, or loss-of-function mutations in tumor suppressor genes, which result in underexpression or lack of activity of the gene product. The protein products of oncogenes having gain-of-function mutations are technically difficult drug targets, due to the lack of effective strategies to selectively inhibit solely the excessive activity of the protein in tumor cells, without deleteriously affecting necessary levels of protein activity in normal cells. Conversely, tumor suppressor genes with loss-of-function mutations are also problematic as drug targets, as it is technically very difficult to develop small molecule drugs that restore the function of a missing or defective protein.
Thus, there is a need for systematic methods to identify highly selective drugs and their cognate targets for killing or inhibiting the proliferation of cancer cells by exploiting the specific genetic alterations that characterize tumor cells. Genetic screening in model organisms offers one possible solution to this challenge. Large-scale, systematic genetic screens in model organisms provide a technically feasible strategy for functionally analyzing nearly all genes and gene products within an organism that relate to a physiological process of interest, and are robust and efficient enough to identify extremely rare genetic mutations. This approach has been used routinely to dissect physiologically important pathways in a number of genetically facile species including the baker""s yeast Saccharomyces cerevisiae, the nematode Caenorhabditis elegans, the fruit fly Drosophila melanogaster, the zebrafish Danio rerio, and the mouse Mus musculus. With respect to using these model organisms for analyzing processes that relate to human disease, each model organism has its own advantages and disadvantages which generally reflect a balance of technical ease of manipulation versus direct relevance to human genetics and physiology. Factors affecting technical ease of use in each system include generation time, cost of growth and maintenance, genome size, and availability of tools for genetic engineering, mutagenesis, gene mapping, and gene cloning. Consequently, the unicellular yeast S. cerevisiae offers perhaps the greatest technical facility for genetic screens with a short generation time of only 2 hours, a haploid phase of the life cycle, and a small genome size less than {fraction (1/100)} that of human; however, this system suffers from the fact that baker""s yeast is a unicellular organism and many genes and pathways involved in intracellular communication, differentiation, and growth control in humans are completely absent in S. cerevisiae. Conversely, as a mammal the mouse is clearly the most similar model organism in genome organization and physiology to human, but suffers from that fact that growth, maintenance, and manipulation of mice is relatively cumbersome, time consuming and expensive. Accordingly, the invertebrate animal model organisms, C. elegans and Drosophila, have found favor for large scale genetic screens because they have provided an especially effective comprise between ease of manipulation and functional relevance to human physiology.
Beyond the issues of the technical feasibility of performing large scale genetic screens with model organisms, properly designed genetic screening strategies provide an efficient and logically rigorous method to identify ideal drug targets. Most drugs act by specifically inhibiting the activity of the target proteins with which they associate. And, most mutations generated by mutagenesis in genetic screens are loss-of-function mutations which reduce the expression or activity of the protein products of those genes. Thus, it follows that a loss-of-function mutation can be considered a surrogate for the effect of a drug that specifically inhibits the activity of the protein product of that gene; and further a loss-of-function mutation in a gene which produces a phenotype in vivo that mimics a desired therapeutic effect therefore identifies as a potential drug target the protein product of that mutant gene. So, the challenge in using genetic screens to identify novel drug targets for a particular disease is to carefully design the screen such that the desired loss-of-function mutations, which simulate the ideal therapeutic effect of a drug in vivo, can be readily and efficiently selected by virtue of a specific, easily scored phenotype.
In fact, large scale genetic screens in model organisms have been extensively employed to dissect genetic and biochemical pathways that relate to fundamental aspects of cancer biology. For example, genetic analysis of yeast has proven to be a very valuable approach to identify genes and proteins involved in DNA repair (Friedberg, Micrbiol Rev (1988) 52:70) and control of the cell cycle (Hartwell, J. Cell Biol. (1980) 85:811-822). Genetic analysis in the nematode C. elegans has led to important discoveries regarding growth factor signaling, for example through the ras pathway (Kayne and Stemberg, Curr Opin Genet Dev (1995) 5:38-43), and factors involved in controlling apoptosis (Ellis and Horvitz, Cell (1986) 44:817-829). Similarly, large scale genetic screens in the fruit fly Drosophila have also led to the discovery of novel components of cancer associated signal transduction pathways, including the ras (Karim et al., Genetics (1996) 143:315-329), notch (Go and Artavanis-Tsakonas, Genetics (1998) 150:211-220), dpp (Raftery et al., Genetics (1995) 139:241-254), and hedgehog (Hooper and Scott, Cell (1989) 59:751-765) pathways. Indeed, gene mutations in Drosophila that affect tumor formation in this simple invertebrate organism were identified as long as 80 years ago (Stark, J. Cancer Res. (1918) 3:279-299; Gateff and Schneiderman, Natl. Cancer Inst Monogr (1969) 31:365-397). The physiological relevance of Drosophila tumors to those of mammals has been convincingly demonstrated by studies of the lats tumor suppressor gene. Mutations in the lats gene in Drosophila were first discovered in genetic screens specifically designed to identify tumor suppressor-like genes in this organism (Xu et al., Development (1995) 117:1223-1237). In these studies, genetically mosaic Drosophila were generated containing clones of cells with homozygous mutations in the lats gene, and these clones of mutant cells were found to develop into large tumors. Most importantly, human and mouse homologs of the Drosophila lats gene were identified, and knockout mice genetically engineered to contain homozygous mutations in the Lats gene developed soft-tissue sarcomas, ovarian stromal cell tumors, hyperplastic changes in the pituitary, and were highly sensitive to carcinogenic treatments (St. John et al., Nat Genet (1999) 21:182-186) thereby validating the physiological relevance of tumor suppressor genes in Drosophila with counterparts in mammals. The technical obstacles posed by mammalian model systems such as the mouse have made it more costly and less practical to pursue large scale mutagenesis schemes by comparison with the simpler invertebrate organisms; nonetheless, recent progress has provided tools to perform in vivo mutagenesis in the mouse efficiently enough to enable phenotype-based genetic screens (McDonald, Proc Soc Exp Biol Med (1995) 209:303-308; Beddington, Curr Biol (1998) 8:R840-R842; Schimenti and Buchan, Genome (1998) Res 8:698-710). With respect to applications to cancer, a genetic screen in the mouse using the chemical mutagen ethylnitrosourea was used to identify a dominant mutation that predisposes to spontaneous intestinal cancer (Moser et al., Science (1990) 247:322-324). Subsequent characterization of this particular mutation revealed that the mutant mice contained a genetic lesion in the murine homolog of the human APC gene, a gene mutated in patients afflicted with familial adenomatous polyposis which predisposes to colorectal cancer, and which is also mutated in sporadic colorectal cancers (Su et al., Science (1992) 256: 668-670). As a further variation beyond ordinary phenotype-based forward genetic screening in the mouse, methods for genetic modifier screens have been proposed based on the identification of mutations that enhance or suppress a starting index phenotype of interest, such as modification of the predisposition to intestinal cancer caused by mutations in the murine homolog of the APC gene (U.S. Pat. No. 5,780,236).
Although all of the genetic screening approaches in model organisms described above have proven their utility in the context of basic research, it is significant to note that the general object of such screens is to systematically identify as many components as possible of biological pathways of interest, for example those biological pathways linked to cancer. Importantly, such screens are not designed to selectively identify genes encoding proteins having properties expected for ideal anti-cancer drug targets. Thus, much time and effort must be invested to carry out functional analysis and validation on the genes arising from such screens to identify those few which are most promising as practical drug targets. More recently, a type of genetic screen directly aimed at identifying drug targets for the treatment of cancer has been described (PCT WO9924603; Hartwell et al., Science (1997) 278:1064-1068; Friend and Oliff, New Eng. J. Med. (1998) 338:125-126). This directed genetic screen has been most well-defined in its application in the yeast S. cerevisiae, and is actually based on a well-established method in yeast genetics termed a xe2x80x9csynthetic lethal screenxe2x80x9d which has been commonly employed as a tool to investigate redundant genetic functions in yeast (Botstein, Harvey Lectures (1987) 82:157-167; Doye and Hurt, Trends Genet (1995) 11:235-241). As a directed strategy to identify anti-cancer drug targets using yeast genetics, the standard synthetic lethal screening strategy has been modified slightly by starting with a yeast strain that contains a primary mutation in a yeast gene that is homologous to a human or mammalian gene mutated in cancer cells, for example a mutant yeast strain modified in the DNA repair genes MLH1 or MSH2 (CT WO9924603; Hartwell et al., supra). As a consequence, this variation of a synthetic lethal genetic screen specifically selects for secondary mutations which result in a lethal phenotype in yeast only in the context of a primary mutation within a yeast homolog of a human cancer-associated gene (such as MLH1 or MSH2), but not in the context of the wild-type form of the cancer-associated gene.
The underlying logic of using a synthetic lethal screen in yeast to identify anti-cancer drug targets is sound, but it poses serious limitations with respect to ultimate applicability to human cancer. In particular, homologs of many of the genes and pathways that are the most commonly mutated in human cancer are completely missing in yeast and are beyond the scope of these screens- including Rb, p53, growth factor signaling, and apoptosis pathways. It is these genes that clearly represent the most desirable starting points for such a screen as the resulting drug targets would have the broadest clinical applicability across different forms of cancer. The absence of these most common cancer associated genes and pathways in yeast reflects that fact that S. cerevisiae is a very simple unicellular organism. This leads to the question of whether the methods of the yeast synthetic lethal screen to identify anti-cancer drug targets can be directly transferred to a more complex multicellular metazoan animal, such as the nematode C. elegans, the fruit fly Drosophila or the mouse, where the common cancer-associated pathways are present. Unfortunately, this is not feasible as the yeast genetic screening method relies on plasmid manipulation techniques that are not applicable in metazoans. Also, the screening methods in yeast take advantage of the haploid life cycle in this organism, where there is just a single copy of each yeast gene, to efficiently identify genes where loss-of-function mutations would otherwise be recessive.
Although synthetic lethal genetic interactions have long been known in metazoan animal genetics (Dobzhansky, Genetics (1946) 31:261-290), a genetic screen based on identification of an organismal synthetic lethal phenotype would be extremely inefficient in metazoans, due to the inability to directly recover the secondary mutation from the dead animals for further analysis and characterization. As a result, other forms of genetic screens to selectively identify anti-cancer drug targets have been proposed for use in metazoans as substitutes for a true synthetic lethal screening strategy. Specifically, it has been suggested that a genetic modifier screen in Drosophila based on identifying mutations that act as enhancers of a rough eye phenotype induced by ectopic expression of dmyc (a homolog of the human myc oncogene) would be the xe2x80x9cconceptual equivalentxe2x80x9d of synthetic lethal mutations; hence, the genes having the enhancer mutations would have the properties of ideal anti-cancer drug targets (WO 99/24603). Actually, this is not the case; enhancer mutations arising from such a dmyc misexpression modifier screen would generally identify genes whose normal role is to down-regulate the activity of the dmyc pathway, and these enhancer mutations would not necessarily exhibit the desired property of a mutation in a gene encoding an ideal anti-cancer drug target, i.e. specifically killing or inhibiting the proliferation of only those cells were the dmyc oncogene is abnormally overexpressed.
In view of the foregoing, it is apparent that improved methods for identifying anti-cancer drug targets are highly desirable.
The invention provides highly selective methods for the identification of anti-cancer drug targets. One method comprises recombinantly modifying a non-human metazoan animal to mis-express a tumor gene or an oncogene (collectively referred to as xe2x80x9csensitizer genesxe2x80x9d) in a target tissue. Preferably the non-human animal is one that is commonly used in genetic studies (e.g. Drosophila, C. elegans, or mouse). The target tissue is dispensable for viability and reproduction of the animal. In some embodiments, mis-expression of the sensitizer gene results in abnormal proliferation of the target tissue, while non-targeted tissues of the recombinantly modified animal exhibit normal proliferation. Progeny of the recombinantly modified animal are generated that have mutations in putative xe2x80x9cinteractorxe2x80x9d genes that may result in sensitizer gene-specific antiproliferation or killing of the target tissue. The interactor gene mutations may be specifically targeted, for example using RNA interference to knock out gene function, or may be randomly generated using chemical, radiation, or transposon mutagenesis. Progeny are identified that have a reduction in size or absence of the target tissue, but have normal non-target tissues. These animals, which are said to exhibit sensitizer gene-specific antiproliferation phenotypes, are further evaluated to determine whether they have mutatations in interactor genes that cause the phenotype. Interactor genes and homologues and orthologs thereof, are isolated and used to identify anti-tumor compounds.
In an alternative embodiment of the invention, the recombinantly modified animals and their progeny are used to directly screen anti-tumor compounds. Test compounds are administered to the animals in varying doses. Compounds that cause a reduction in size of the target tissue and have no adverse effects on non-target tissues are further evaluated as putative anti-tumor compounds.
The invention also provides various novel transformation constructs that can be used to generate novel transgenic animals that have altered expression of a sensitizer gene in a target tissue. The constructs are diagrammed in FIGS. 2, 6, and 8, and are described in detail in the specification.
In another alternative embodiment of the invention, RNA interference (RNAi) is used in cell culture-based screens to identify interactor genes that exhibit sensitizer gene-specific antiproliferation phenotypes, or as confirmation screens of interactor genes identified in the progeny of the recombinantly modified animals.
The entire contents of all references, including patent applications, cited herein are incorporated by reference in their entireties for all purposes. Additionally, the citation of a reference in the preceding background section is not an admission of prior art against the invention described herein.